Method for analyzing the effect of sediment particle size on salinity measurements based on gbrt regression trees

Abstract

The application discloses a method for analyzing the influence of sediment particle size on salinity measurement based on a GBRT regression tree, which comprises the following steps: preparing a sediment sample, wherein the sediment sample comprises a plurality of sand sample groups in different particle size intervals; confirming the minimum equation number; processing the plurality of sand sample groups according to the minimum equation number to obtain a corresponding number of new sand sample groups; performing particle size analysis and salinity detection on each new sand sample group; and training a fine particle sediment group influence analysis model based on the GBRT regression tree by using the particle size analysis result and the salinity detection data to obtain an estimation function. The method for analyzing the influence of sediment particle size on salinity measurement based on the GBRT regression tree provided by the application solves the influence of fine sediment particle size on the salinity measurement based on the conductivity by using an intelligent algorithm.

Description

Technical Field

[0001] This invention belongs to the field of seawater salinity measurement technology, specifically relating to an analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression trees. Background Technology

[0002] Salinity is the most fundamental parameter in oceanography. Many phenomena in the ocean are closely related to the distribution and changes in salinity, and salinity data are of great significance for oceanographic research, scientific experiments, and engineering practice.

[0003] Silver nitrate titration is the only practical method for determining the exact grams of dissolved salt in seawater; however, it is time-consuming and complex. Methods such as specific gravity, refraction, and acoustic methods have low accuracy in salinity measurement and cannot meet the needs of modern oceanographic measurements. Satellite remote sensing can rapidly acquire large-area or globally synchronous continuous ocean surface salinity data, primarily used in global water cycles and ocean currents. Conductivity-based salinity measurement is simple to operate, fast, and accurate, making it a commonly used method for ocean salinity observation. The 1978 Practical Salt Standard, based on conductivity, is widely used in marine surveys and salinity measurements in estuaries and coastlines. The standard seawater for the 1978 Practical Salt Standard was taken from the deep North Atlantic Ocean; therefore, the practical salinity definition did not consider the influence of suspended sediment. Without considering clay minerals, organic matter, etc., the influence of suspended sediment itself on salinity measurement includes suspended sediment concentration and particle size. Summary of the Invention

[0004] This invention provides an analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression trees to solve the aforementioned technical problems. Specifically, the technical solution is as follows:

[0005] An analytical method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression trees, comprising:

[0006] S1: Prepare sediment samples, wherein the sediment samples include several groups of sand samples with different particle size ranges;

[0007] S2: Determine the minimum number of equations N max ;

[0008] S3: Based on the minimum number of equations N max Processing several of the aforementioned sand sample groups yields a corresponding number of new sand sample groups;

[0009] S4: Perform particle size analysis and salinity detection on each of the new sand sample groups;

[0010] S5: The estimated function is obtained by training a fine-grained sediment group influence analysis model based on GBRT regression tree using particle size analysis results and salinity detection data.

[0011] Further, step S1 includes:

[0012] S11: Collect raw sand samples and perform pretreatment;

[0013] S12: The pretreated sand sample was extracted using an intelligent water analysis device to obtain seven suspensions with particle size ranges of 0-10μm, 0-10μm, 10-20μm, 20-30μm, 30-40μm, 40-50μm, 50-60μm, and 60-70μm. The sand sample groups were prepared by drying and crushing, and were designated as Sed1, Sed2, Sed3, Sed4, Sed5, Sed6, and Sed7.

[0014] Furthermore, in step S12, the pretreated sand sample is extracted using an intelligent water analysis device.

[0015] The intelligent water analysis device includes:

[0016] The container has an opening at the top, seven slits spaced apart on one side of the container, sealing strips provided on the upper and lower sides of the slits, and seven discharge holes spaced apart on the other side of the container. Each discharge hole is connected to a discharge pipe, and each discharge pipe is equipped with a suction pump and a valve.

[0017] Seven partitions are inserted into the seven gaps to divide the container into several isolated spaces. The height of the spaces satisfies H21:H22:H23:H24:H25:H26:H27 = 1:4:9:16:25:36:49. Each discharge hole is connected to a corresponding isolated space. Each partition includes a plate body and a handle connected to one end of the plate body.

[0018] A pull assembly is connected to a handle on the partition; the drag assembly is equipped with a motor and drives the partition via the motor.

[0019] The control unit is connected to the motor and the suction pump.

[0020] Furthermore, magnets are embedded in both the sealing strip and the handle.

[0021] Furthermore, each of the partition plates has an arc-shaped notch at a corresponding position on its side.

[0022] The intelligent water analysis device includes:

[0023] An ultrasonic vibrating rod is inserted into the container by passing through each of the notches in sequence.

[0024] Further, step S2 includes:

[0025] S21: Mix the sand sample group in the smallest particle size range and the sand ocean group in the largest particle size range to obtain a mixed sand sample group;

[0026] S22: Analyze the particle size distribution data of the mixed sand sample group using a laser particle size analyzer. The particle size distribution data includes the percentage content corresponding to different effective particle size groups, and the number of effective particle size groups in the mixed sand sample group is used as the minimum equation number N. max .

[0027] Further, step S4 includes:

[0028] S41: Analyze the particle size distribution data of each of the new sand sample groups using a laser particle size analyzer. The particle size distribution data includes the percentage content corresponding to different effective particle size groups.

[0029] S42: Measure the salinity of each of the new sand sample groups using a salinity measuring device.

[0030] Further, step S5 includes:

[0031] S51: Construct a structure containing N max A regression model of a system of equations,

[0032]

[0033] Where, x n,1 x n,2 ... x n,Nmax The percentage content corresponding to different effective particle size groups, c1, c2, ..., c Nmax The coefficients representing the influence of different effective particle size groups on salinity are y1, y2, ..., y Nmax To measure salinity;

[0034] S52: The influence coefficient is calculated using the GBRT regression tree algorithm.

[0035] Further, step S52 includes:

[0036] Step S521: Given a training sample set,

[0037]

[0038] Output estimation function f(x) n,i );

[0039] Step S522: Define the linear regression equation:

[0040] f(x) = cx + b

[0041] Where, x = (x n,1 ,x n,2 ,L,x n,Nmax) l l is an n-dimensional unit column vector, c is the influence coefficient of different effective particle size groups on salinity, and b is a constant.

[0042] Step S523: Find the optimal hyperplane. The objective function at this point is:

[0043]

[0044] sty-xc-bI=ξ

[0045] Construct the Lagrange function:

[0046]

[0047] Among them, γ, ξ, α, and I are auxiliary variables set in the vector machine model and have no actual physical meaning;

[0048] Step S524: Take the partial derivatives of the Lagrange function with respect to c, b, ξ, and α respectively, and set them to 0 to obtain the estimation function of the least squares support vector machine.

[0049]

[0050] Furthermore, the analytical method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression trees also includes:

[0051] After measuring the particle size distribution and percentage content of the sediment to be analyzed, the salinity under the influence of fine-particle sediment is obtained by substituting the values ​​into the estimation function.

[0052] The advantage of this invention lies in the analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression tree, which solves the problem of the influence of fine sediment particle size on conductivity-based salinity measurement through intelligent algorithms.

[0053] The advantages of this invention lie in the analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression trees. The intelligent water analysis device used controls the settling time to automate the movement of the baffle and the suction of the suspension, thus improving the efficiency of sand sample preparation. Simultaneously, the ultrasonic vibrator helps disperse fine sediment flocs, improving the accuracy of sediment water analysis. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of an analytical method for analyzing the influence of sediment particle size on salinity measurement based on a GBRT regression tree according to the present invention.

[0056] Figure 2 This is a schematic diagram of the intelligent water analysis device of the present invention;

[0057] Figure 3 This is a schematic diagram of the intelligent water analysis device of the present invention;

[0058] Figure 4 This is a schematic diagram of the partition spacing of the intelligent water analysis device of the present invention;

[0059] Figure 5 This is a schematic diagram of the salinity measuring device of the present invention. Detailed Implementation

[0060] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0061] like Figure 1 The diagram illustrates an analytical method based on a GBRT regression tree for analyzing the influence of sediment particle size on salinity measurement, comprising: S1: preparing sediment samples, wherein the sediment samples include several groups of sand samples with different particle size ranges. S2: confirming the minimum equation number N. max S3: Based on the minimum number of equations N max Several sand sample groups are processed to obtain a corresponding number of new sand sample groups. S4: Particle size analysis and salinity detection are performed on each new sand sample group. S5: The influence analysis model of fine-particle sediment groups based on GBRT regression tree is trained using the particle size analysis results and salinity detection data to obtain the estimation function. The method for analyzing the influence of sediment particle size on salinity measurement based on GBRT (GradientBoost Regression Tree) in this application solves the influence of refined sediment particle size on conductivity-based salinity measurement through intelligent algorithms. The above steps are described in detail below.

[0062] For step S1: preparing sediment samples, the sediment samples include several groups of sand samples with different particle size ranges.

[0063] In a preferred embodiment, step S1 includes:

[0064] S11: Collect raw sand samples and perform pretreatment. Specifically, collect raw sand samples from the upper reaches of the tidal boundary of the estuary to reduce the impact of sediment-borne salinity on salinity detection; the raw sand samples may contain debris such as boulders and tree roots, and are initially screened using a wire mesh with a 100μm aperture.

[0065] S12: The pretreated sand sample was extracted using the static water sedimentation method of the intelligent water analysis device 10 to obtain seven suspensions with particle size ranges of 0-10μm, 0-10μm, 10-20μm, 20-30μm, 30-40μm, 40-50μm, 50-60μm, and 60-70μm. The sand sample groups were prepared by drying and crushing, and were designated as Sed1, Sed2, Sed3, Sed4, Sed5, Sed6, and Sed7.

[0066] The intelligent water analysis device 10 uses a static water sedimentation method to extract sand samples. Specifically, such as... Figure 2-4 As shown, the intelligent water analysis device 10 includes: a container 18, seven partitions 20, a pull assembly, and a control unit 32.

[0067] Specifically, container 18 is a rectangular or cylindrical container with an open top, preferably a rectangular shape. The material can be glass, iron, or plastic. For observing the sedimentation of sediment inside container 18, acrylic is preferably used. To separate sand sample groups (Sed1 to Sed7), seven slits 11 are spaced apart on one side 16 of container 18. The length of each slit 11 is consistent with the width 12 (H1) of the side wall of container 18 and the width 12 of the partition 20. Sealing strips 13 are provided on the upper and lower sides of each slit 11. Magnets are embedded in the sealing strips 13. Seven partitions 20 are inserted into the seven slits 11 to divide container 18 into several isolated spaces. The partitions 20 are used to separate the test sand sample after a period of time during the sedimentation of the original sand sample in purified water. The partition 20 includes a plate 21 and a handle 22 connected to one end of the plate 21. The plate 21 is preferably an acrylic plate. A magnet is embedded in the handle 22. When the partition 20 is fully inserted, the handle 22 and the sealing strip 13 attract each other, increasing the sealing performance.

[0068] On the other side of container 18, seven discharge holes 15 are spaced apart, each discharge hole 15 connecting to a corresponding isolation space. Each discharge hole 15 is connected to a discharge pipe 41, and each discharge pipe 41 is equipped with a suction pump 42 and a valve 43. Furthermore, each discharge pipe 41 is connected to a collection pool 44 for storing test sand samples (Sed1 to Sed7).

[0069] The pull assembly is connected to the handle of the partition 20, and the drag assembly is equipped with a motor 31, which drives the partition 20. Specifically, the pull assembly includes a pull rod 33, a guide rod 34 and a motor 31, which inserts or pulls the partition 20 through the pull rod 33.

[0070] The control unit 32 is connected to the motor 31 and the suction pump 42. The control unit 32 can synchronously control the levers 33 of all partitions 20 or individually control a single lever 33, and can also synchronously control all suction pumps 42 or individually control a single suction pump 42. In this application, the control unit 32 is a computer.

[0071] In a preferred embodiment, each partition 20 has an arc-shaped notch 23 on the side of its plate 21. The intelligent water separation device 10 includes an ultrasonic vibrating rod 14. The ultrasonic vibrating rod 14 passes through each notch 23 sequentially to be inserted into the container 18. Because fine-particle sediment is sticky, it easily flocculates and agglomerates during still water settling, increasing particle size and causing fine-particle sediment to enter the coarse particle sequence, affecting the screening effect. Therefore, this invention provides an ultrasonic vibrating rod 14 to disperse fine-particle flocculations by applying ultrasonic waves to the water, thus keeping the fine-particle sediment in a discrete state and solving the above problem. The ultrasonic vibrating rod 14 is connected to an ultrasonic vibrator (not shown) via a wire 17.

[0072] like Figure 4 As shown, due to the different particle sizes, the settling distances are different in the same amount of time. Therefore, according to the Stockes settling velocity formula, the settling velocities at 10μm, 20μm, 30μm, 40μm, 50μm, 60μm, and 70μm can be calculated, and the settling distances in the same amount of time can be calculated.

[0073]

[0074] Where: w is the settling velocity of sediment particles; ρ s ρ is the density of sediment particles; w ρ is the density of the water; g is the acceleration due to gravity; R is the radius of the sediment particles; μ is the viscosity of the water.

[0075] According to the calculation results in Table 1, the heights (spacing between partitions) of several spaces satisfy H21:H22:H23:H24:H25:H26:H27 = 1:4:9:16:25:36:49. With this setup, after the same amount of time, the suspensions of Sed1 concentrate in space 181, Sed2 in space 182, Sed3 in space 183, Sed4 in space 184, Sed5 in space 185, Sed6 in space 186, and Sed7 in space 187; sediment particles larger than 70 μm enter space 188.

[0076] Table 1. Calculation of diaphragm spacing based on Stockes' sinking velocity formula.

[0077]

[0078] For step S2: Determine the minimum number of equations N max .

[0079] In a preferred embodiment, step S2 includes:

[0080] S21: Mix the sand sample group with the smallest particle size range and the sand sample group with the largest particle size range to obtain a mixed sand sample group.

[0081] S22: Analyze the particle size distribution data of the mixed sand sample group using a laser particle size analyzer. The particle size distribution data includes the percentage content corresponding to different effective particle size groups, and the number of effective particle size groups in the mixed sand sample group is used as the minimum equation number N. max .

[0082] In this application, a laser particle size analyzer was used to analyze the particle size of each group of sediment. Taking the Sed1 (0–10 μm) test sand sample as an example, as shown in Table 2, the percentage content of sediment with 84 different particle sizes can be obtained, of which 46 are effective particle size groups, denoted as N1. Similarly, for the test sand samples (Sed2–Sed7), the effective particle size groups are denoted as N2–N7 respectively. Since N1–N7 alone cannot cover the smallest to largest particle size range, Sed1 and Sed7 are mixed, and the effective particle size group is denoted as N. max Therefore, the number of variables to be solved is N. max The minimum number of equations required is N. max .

[0083] Table 2. Particle size analysis of sand sample Sed1

[0084] Particle size (µm) Interval % accumulation% Particle size (µm) Interval % accumulation% Particle size (µm) Interval % accumulation% Particle size (µm) Interval % accumulation% 0.000-0.100 0 0 0.970-1.087 2.07 14.76 10.56-11.83 4.23 81.12 114.9-128.7 0 100 0.100-0.112 0 0 1.087-1.218 2.22 16.98 11.83-13.25 3.9 85.02 128.7-144.2 0 100 0.112-0.125 0 0 1.218-1.365 2.35 19.33 13.25-14.85 3.54 88.56 144.2-161.5 0 100 0.125-0.140 0 0 1.365-1.529 2.43 21.76 14.85-16.64 3.12 91.68 161.5-181.0 0 100 0.140-0.157 0 0 1.529-1.713 2.47 24.23 16.64-18.64 2.65 94.33 181.0-202.8 0 100 0.157-0.176 0 0 1.713-1.920 2.44 26.67 18.64-20.88 2.12 96.45 202.8-227.2 0 100 0.176-0.197 0 0 1.920-2.151 2.44 29.11 20.88-23.40 1.58 98.03 227.2-254.6 0 100 0.197-0.221 0 0 2.151-2.410 2.47 31.58 23.40-26.22 1.04 99.07 254.6-285.2 0 100 0.221-0.248 0.01 0.01 2.410-2.700 2.58 34.16 26.22-29.37 0.58 99.65 285.2-319.5 0 100 0.248-0.278 0.05 0.06 2.700-3.025 2.69 36.85 29.37-32.91 0.25 99.9 319.5-358.0 0 100 0.278-0.311 0.17 0.23 3.025-3.389 2.82 39.67 32.91-36.87 0.08 99.98 358.0-401.1 0 100 0.311-0.349 0.36 0.59 3.389-3.797 2.94 42.61 36.87-41.31 0.02 100 401.1-449.4 0 100 0.349-0.391 0.54 1.13 3.797-4.254 3.06 45.67 41.31-46.28 0 100 449.4-503.5 0 100 0.391-0.438 0.75 1.88 4.254-4.767 3.2 48.87 46.28-51.85 0 100 503.5-564.1 0 100 0.438-0.490 0.97 2.85 4.767-5.340 3.34 52.21 51.85-58.10 0 100 564.1-632.0 0 100 0.490-0.550 1.24 4.09 5.340-5.983 3.54 55.75 58.10-65.09 0 100 632.0-708.1 0 100 0.550-0.616 1.43 5.52 5.983-6.703 3.79 59.54 65.09-72.92 0 100 708.1-793.3 0 100 0.616-0.690 1.62 7.14 6.703-7.510 4.07 63.61 72.92-81.70 0 100 793.3-888.8 0 100 0.690-0.773 1.75 8.89 7.510-8.414 4.33 67.94 81.70-91.54 0 100 888.8-995.8 0 100 0.773-0.866 1.86 10.75 8.414-9.427 4.49 72.43 91.54-102.5 0 100 995.8-1115 0 100 0.866-0.970 1.94 12.69 9.427-10.56 4.46 76.89 102.5-114.9 0 100 1115-1250 0 100

[0085] For step S3: Based on the minimum number of equations N max Several sand sample groups were processed to obtain a corresponding number of new sand sample groups.

[0086] Specifically, since there are only seven groups of sand samples in a single experiment (Sed1~Sed7), seven systems of equations can be established, which is less than the minimum number of equations N. max The solution cannot be found.

[0087] Therefore, by mixing sediments of different particle sizes, such as mixing Sed1-Sed2, Sed1-Sed3, or mixing Sed1 with Sed2 and Sed3, or Sed2 with Sed3 and Sed4, the final result satisfies the requirement that the number of experimental sand sample groups satisfies the minimum number of equations N. max Requirements.

[0088] The influence of experimental sand samples (Sed1–Sed7) on salinity measurements was investigated under the same water salinity and sediment concentration. Figure 5As shown, the salinity measuring device 50 mainly includes an iron stand 51, a laptop computer 52, a thermostatic magnetic stirrer 53, a CTD measuring instrument 54, and a beaker 55. The beaker 55 is used to store the test sand sample suspension 56, and the rotor 57 of the thermostatic magnetic stirrer 53 maintains the mixing uniformity of the suspension 56. The CTD measuring instrument 54 is immersed in the suspension 46 to measure salinity, and the salinity measurement data is transmitted to the laptop computer 52 via a data cable 58. The salinity measuring device 50 has automatic recording and online display functions.

[0089] For step S4: Perform particle size analysis and salinity testing on each new sand sample group.

[0090] In a preferred embodiment, step S4 includes:

[0091] S41: Analyze the particle size distribution data of each new sand sample group using a laser particle size analyzer. The particle size distribution data includes the percentage content corresponding to different effective particle size groups.

[0092] S42: Measure the salinity of each new sand sample group using a salinity measuring device.

[0093] For step S5: The estimated function is obtained by training a fine-grained sediment group impact analysis model based on GBRT regression trees using particle size analysis results and salinity detection data. The GBRT regression tree employs a linear search method in each cell region to ensure the loss function reaches its minimum in each cell region, thereby minimizing the overall loss function. The corresponding return value is then found by optimizing the loss function.

[0094] In a preferred embodiment, step S5 includes:

[0095] S51: Construct a structure containing N max A regression model of a system of equations,

[0096]

[0097] Where, x n,1 x n,2 ... x n,Nmax The percentage content corresponding to different effective particle size groups, c1, c2, ..., c Nmax The coefficients representing the influence of different effective particle size groups on salinity are y1, y2, ..., y Nmax To measure salinity.

[0098] S52: The influence coefficient is calculated using the GBRT regression tree algorithm.

[0099] In a preferred embodiment, step S52 includes:

[0100] Step S521: Given a training sample set,

[0101]

[0102] Output estimation function f(x) n,i ).

[0103] Step S522: Define the linear regression equation:

[0104] f(x) = cx + b

[0105] Where, x = (x n,1 ,x n,2 ,L,x n,Nmax ) l l is an n-dimensional unit column vector, c is the influence coefficient of different effective particle size groups on salinity, and b is a constant.

[0106] Step S523: Find the optimal hyperplane. The objective function at this point is:

[0107]

[0108] sty-xc-bI=ξ

[0109] Construct the Lagrange function:

[0110]

[0111] Here, γ, ξ, α, and I are auxiliary variables set in the vector machine model and have no actual physical meaning.

[0112] Step S524: Take the partial derivatives of the Lagrange function with respect to c, b, ξ, and α respectively, and set them to 0 to obtain the estimation function of the least squares support vector machine.

[0113]

[0114] As a preferred implementation method, the analysis method for the influence of sediment particle size on salinity measurement based on GBRT regression tree also includes: measuring the particle size distribution and percentage content of the sediment to be analyzed and then substituting them into an estimation function to obtain the measured salinity under the influence of fine-particle sediment.

[0115] Understandably, when it is necessary to analyze the impact of fine-grained sediment on salinity measurement, it is only necessary to use a laser particle size analyzer to measure the sediment particle size distribution and percentage content (x). i Substitution The salinity f(x) under the influence of fine-grained sediment can then be calculated. i Compared to the measured salinity f(x0) without added sediment, the absolute value of the salinity difference under the influence of fine-grained sediment can be obtained: Δy=|f(x0) ... i)-f(x0)|.

[0116] In summary, the GBRT regression tree can effectively establish a salinity regression model under the synergistic influence of multiple factors on fine-grained sediment.

[0117] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any way, and all technical solutions obtained by equivalent substitution or equivalent transformation fall within the protection scope of the present invention.

Claims

1. An analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression trees, characterized in that, Include: S1: Prepare sediment samples, wherein the sediment samples include several groups of sand samples with different particle size ranges; S2: Determine the minimum number of equations N max ; S3: Based on the minimum number of equations N max Processing several of the aforementioned sand sample groups yields a corresponding number of new sand sample groups; S4: Perform particle size analysis and salinity detection on each of the new sand sample groups; S5: The estimation function is obtained by training a fine-grained sediment group influence analysis model based on GBRT regression tree using particle size analysis results and salinity detection data; Step S2 includes: S21: Mix the sand sample group in the smallest particle size range and the sand ocean group in the largest particle size range to obtain a mixed sand sample group; S22: Analyze the particle size distribution data of the mixed sand sample group using a laser particle size analyzer. The particle size distribution data includes the percentage content corresponding to different effective particle size groups, and the number of effective particle size groups in the mixed sand sample group is used as the minimum equation number N. max .

2. The analytical method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 1, characterized in that, Step S1 includes: S11: Collect raw sand samples and perform pretreatment; S12: The pretreated sand sample was extracted using an intelligent water analysis device to obtain seven suspensions with particle size ranges of 0~10 μm, 10~20 μm, 20~30 μm, 30~40 μm, 40~50 μm, 50~60 μm, and 60~70 μm. The sand sample groups were prepared by drying and crushing, and were designated as Sed1, Sed2, Sed3, Sed4, Sed5, Sed6, and Sed7.

3. The analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 2, characterized in that, In step S12, the pretreated sand sample is extracted using an intelligent water analysis device; The intelligent water analysis device includes: The container has an opening at the top, seven slits spaced apart on one side of the container, sealing strips provided on the upper and lower sides of the slits, and seven discharge holes spaced apart on the other side of the container. Each discharge hole is connected to a discharge pipe, and each discharge pipe is equipped with a suction pump and a valve. Seven partitions are inserted into the seven gaps to divide the container into several isolated spaces. The height of the spaces satisfies H21:H22:H23:H24:H25:H26:H27=1:4:9:16:25:36:

49. Each discharge hole is connected to a corresponding isolated space. Each partition includes a plate body and a handle connected to one end of the plate body. A pull assembly is connected to a handle on the partition, the pull assembly being equipped with a motor, and the partition being driven by the motor; The control unit is connected to the motor and the suction pump.

4. The analytical method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 3, characterized in that... Magnets are embedded in both the sealing strip and the handle.

5. The method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 3, characterized in that, Each of the partition plates has a corresponding position on its side panel; The intelligent water analysis device includes: an arc-shaped notch. An ultrasonic vibrating rod is inserted into the container by passing through each of the notches in sequence.

6. The analytical method for the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 1, characterized in that, Step S4 includes: S41: Analyze the particle size distribution data of each of the new sand sample groups using a laser particle size analyzer. The particle size distribution data includes the percentage content corresponding to different effective particle size groups. S42: Measure the salinity of each of the new sand sample groups using a salinity measuring device.

7. The analytical method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 6, characterized in that, Step S5 includes: S51: Construct a structure containing N max A regression model of a system of equations, in, , ... The percentage content corresponding to different effective particle size groups. , ... The effect coefficient of different effective particle size groups on salinity. , ... To measure salinity; S52: The influence coefficient is calculated using the GBRT regression tree algorithm.

8. The method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 7, characterized in that, Step S52 includes: Step S521: Given a training sample set, Output estimation function ; Step S522: Define the linear regression equation: in, l is an n-dimensional unit column vector, c is the influence coefficient of different effective particle size groups on salinity, and b is a constant. Step S523: Find the optimal hyperplane. The objective function at this point is: Construct the Lagrange function: in, , , I is an auxiliary variable set in the vector machine model and has no actual physical meaning; Step S524: Calculate the Lagrange function separately. , , , By taking the partial derivative of and setting it to zero, we can obtain the estimation function of the least squares support vector machine. 。 9. The method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression tree according to claim 1, characterized in that, The analytical method for analyzing the influence of sediment particle size on salinity measurement based on GBRT regression trees also includes: After measuring the particle size distribution and percentage content of the sediment to be analyzed, the salinity under the influence of fine-particle sediment is obtained by substituting the values ​​into the estimation function.

Images